# Fileset

[Angew Chem Int Ed - 2026 - Cheng - Accelerated Discovery‐to‐Unveiling of High‐Performance and Affordable Ammonia Electrode.pdf](https://mdr.nims.go.jp/filesets/20238fcc-a7b4-4d4e-80e1-00d84b9c34c8/download)

## Creator

Yingying Cheng, [Masaki Takeguchi](https://orcid.org/0000-0002-0282-6020), Abraham Castro Garcia, [Ken Sakaushi](https://orcid.org/0000-0003-4797-9087)

## Rights

[Creative Commons BY-NC-ND Attribution-NonCommercial-NoDerivs 4.0 International](https://creativecommons.org/licenses/by-nc-nd/4.0/)

## Other metadata

[Accelerated Discovery‐to‐Unveiling of High‐Performance and Affordable Ammonia Electrode Process by Human–Machine Collaboration Framework](https://mdr.nims.go.jp/datasets/8966d19a-5aae-4bf3-8add-95f616077439)

## Fulltext

Accelerated Discovery‐to‐Unveiling of High‐Performance and Affordable Ammonia Electrode Process by Human–Machine Collaboration FrameworkAngewandte Chemie International EditionHow to cite: Angew. Chem. Int. Ed. 2026, 65, e2469199doi.org/10.1002/anie.2469199RESEARCH ARTICLEAccelerated Discovery-to-Unveiling of High-Performanceand Affordable Ammonia Electrode Process byHuman–Machine Collaboration FrameworkYingying Cheng Masaki Takeguchi Abraham Castro Garcia Ken SakaushiResearch Center for Energy and Environmental Materials, National Institute for Materials Science, Tsukuba, Ibaraki, JapanCorrespondence: Ken Sakaushi (SAKAUSHI.Ken@nims.go.jp)Received: 21 December 2025 Revised: 17 April 2026 Accepted: 23 April 2026Keywords: dimensionally reduced reaction descriptors | electrochemical nitrate reduction reaction | human–machine collaboration | in-situ spectroscopy |kinetic model analysisABSTRACTThe electrochemical nitrate reduction reaction (eNO3RR) to ammonia (NH3) is a key for producing fuels during interstellartravel and an alternative to Haber−Bosch process. However, the complicated multi-electron/proton transfer electrode processof eNO3RR makes affordable electrocatalyst discovery and its mechanistic understanding challenging. Herein, we established ahuman–machine collaboration framework by employing dimensionally reduced reaction descriptors which enables an accelerateddata-driven discovery-to-unveiling of unconventional and high-performance eNO3RR electrocatalysts with desirable elementchoice. Using the current density difference between nitrite (NO2−) reduction and hydrogen evolution as a descriptor, the optimalFeCoNiCuGa electrocatalyst was identified in a drastically short timeframe. Even compared with Pt or Rh, the FeCoNiCuGaexhibits a higher NH3 production rate of 9.8 mmol mgcat−1 at −0.3 V versus a reversible hydrogen electrode. Furthermore,together with a mechanistic study using rotating ring-disk electrode combined with a new kinetic model, in situ infraredspectroscopy unveiled that the adsorbed NO2− (*NO2−) plays a crucial role in the efficient electrode process: a moderate *NO2−binding accelerates NH3 formation whereas a weak binding leads to unfavorable reactions. Our work demonstrates that acomprehensive human–machine collaboration approach enables an accelerated discovery-to-unveiling of promising electrodeprocesses, providing a feasible way to promote game-changing electrochemical technologies.1UtlMseiniTp©AhIntroductionnveiling microscopic mechanisms of multi-electron/-protonransfer electrode processes (MEPTs) is one of the grand chal-enges in science [1–6]. Even for the well-known archetypeEPTs of the hydrogen evolution reaction (HER) at well-definedingle-crystal electrodes, we still do not fully understand howlectrons and protons are transferred at electrified solid–liquidnterfaces, neither theoretically nor experimentally, nor in combi-ation [7–13]. A lack of a precise view of MEPTs is a key bottleneckn electrocatalyst design, delaying the advancement of next-his is an open access article under the terms of the Creative Commons Attribution-NonCrovided the original work is properly cited, the use is non-commercial and no modificati2026 The Author(s). Angewandte Chemie International Edition published by Wiley-VCHngewandte Chemie International Edition, 2026; 65:e2469199ttps://doi.org/10.1002/anie.2469199generation electrochemical technology. Thus, it is game-changingto develop a methodology to accelerate promising electrocatalystdesign with on-demand elements and spontaneously lead to abetter understanding of the microscopic electrode process. In thiscontext, renewable electricity-driven electrochemical ammonia(green-NH3) synthesis is a typical MEPT that faces the issues.NH3 is a vital chemical that is widely used as a fertilizer, afuel, a feedstock for pharmaceuticals, and an energy carrier onEarth and, in the future on the Moon or Mars [14, 15]. Moreover,NH3 remains a critical chemical during interstellar travels. Asommercial-NoDerivs License, which permits use and distribution in any medium,ons or adaptations are made.GmbH1 of 11https://doi.org/10.1002/anie.2469199https://orcid.org/0000-0002-0282-6020https://orcid.org/0000-0003-4797-9087mailto:SAKAUSHI.Ken@nims.go.jphttp://creativecommons.org/licenses/by-nc-nd/4.0/https://doi.org/10.1002/anie.2469199http://crossmark.crossref.org/dialog/?doi=10.1002%2Fanie.2469199&domain=pdf&date_stamp=2026-05-01s[BeaoilaeNs[N(Tt2ar*birmRbgetfdωotmMtapermhsgApmw3tevpfem2 15213773, 2026, 25, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/anie.2469199 by National Institute For, Wiley Online Library on [18/06/2026]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Cruch, global NH3 demand is expected to exceed 650 Mt by 205016]. Currently, NH3 synthesis relies on the Haber–Bosch (H–) process, which operates at high temperatures and generatesxtra CO2 emissions [17, 18]. The green-NH3 synthesis offers anlternative to the H–B process, attracting tremendous attentionver the past decades because it is based on electrolysis, makingt suitable for operation outside Earth or for deployment in theimited space of interstellar spacecraft. Although N2 is the mostbundant nitrogen source, the high N≡N triple bond dissociationnergy (941 kJ mol−1) [19–22] limits NH3 productivity toward2 electrochemical reduction. Water-soluble nitrate (NO3−) pos-esses a low N═O double bond dissociation energy (204 kJ mol−1)20, 23, 24], enabling favorable reaction kinetics for sustainableH3 synthesis from the electrochemical NO3− reduction reactioneNO3RR).he eNO3RR to NH3 is an 8-electrons/9-protons MEPT involvinghe formation of various intermediates and byproducts [20,5]. During eNO3RR, subtle variations in intermediate bindingnd undesired intermediate stabilization can drastically affecteaction barriers and divert the pathway toward side reactions.NO2− is identified as a key intermediate in NH3 production,ut the specific role of *NO2− in regulating NH3 yields remainsnsufficiently explored. In this study, we employed the rotatinging-disk electrode (RRDE) technique alongside effective kineticodeling to elucidate the complex NO3RR electrode process.RDE is a double-working-electrode configuration [26] that haseen widely employed to investigate MEPTs (for example, oxy-en reduction reactions, ORR) [27–35]. In RRDE configuration,lectrochemical redox reactions occur at the disk electrode whilehe ring electrode detects the intermediate or product diffusingrom the disk. By controlling rotation rate (ω), the plots ofisk to ring current (𝑁 𝐼𝐷𝐼𝑅, N is the collection efficiency) versus−0.5 enables in-depth probing of eNO3RR intermediates andverall pathways [27–35]. As a result, we successfully uncoveredheoretical insights for the rational design of multicomponentetal-oxide electrocatalysts.ulticomponent metal-oxides exhibit high specific surface areas,unable electronic structures, and strong synergistic interactionsmong different metal centers, all of which are advantageous forromoting efficient MEPT pathways [36, 37]. However, such ben-fits come at the cost of an extremely large compositional spaceesulting from the random combinations of multiple elements,aking traditional intuition-driven screening strategies pro-ibitively time-consuming and inefficient. This bottleneck under-cores the urgency of developing high-throughput, knowledge-uided approaches to rapidly identify promising candidates.rtificial intelligence (AI) is a powerful and transformativelatform for electrocatalyst discovery, accelerating screening ofulticomponent electrocatalysts by combining domain expertiseith AI descriptor construction and data-driven learning [38,9]. However, applying AI directly to MEPTs is nontrivial due toheir complex reaction networks. For example, a typical MEPTNO3RR involves over a dozen intermediates and producesarious products (NH3, N2, NO2−, et al), which makes selectivityrediction particularly challenging. Therefore, accurately identi-ying highly selective MEPT electrocatalysts requires AI modelsquipped with meaningful descriptors that are tailored to theechanistic complexity of MEPTs.of 11eatHerein, we successfully bridge the gap between AI platformsand MEPTs by developing a directed simplification of MEPTs,enabling accelerated discovery of target materials and fasterunderstanding of microscopic mechanisms. In this concept, aBayesian optimization (BO)-based human–machine collabora-tion framework (HMC) was employed, which effectively short-ened the screening period for eNO3RR electrocatalysts. Using adimensionally reduced reaction descriptor to provide a directedsimplification of MEPT, the optimal quinary electrocatalyst of Fe,Co, Ni, Cu, and Ga based oxide material (denoted FeCoNiCuGa)was identified after 10 “training-suggestion-experiment” loops(1% trials in 3003 candidates). The NH3 production rate of9.8 mmol mgcat−1 is achieved at −0.3 V versus the reversiblehydrogen potential (RHE). Mechanistic studies including RRDEand in situ electrochemical infrared analysis reveal that highNH3 yields originate from the rapid *NO2− formation and con-version, providing guidance for the design of selective eNO3RRelectrocatalysts.2 Results and Discussion2.1 Descriptor EstablishmentGiven the complexity of MEPTs, establishing accurate descriptorsrequires a comprehensive analysis of the overall reaction scheme.The eNO3RR to NH3 is considered a consecutive MEPT, con-sisting of NO3RR to NO2− and electrochemical nitrite reductionreaction (eNO2RR) to NH3, with HER as a side reaction. Inour approach, 15 elements (Cu, Fe, Mn, Sn, Bi, In, Zn, Ag,Au, Co, Ni, Ga, Mo, Sb, and Zr) were selected to constructquinary electrocatalysts. According to previous studies, Cu isresponsible for NO2− formation [40], Fe and Co contribute toNO3− activation [41–43], Sn, Bi, In, Zn, Ag, Au, Ni act as adsorbedH (*H) mediator [44, 45], Ga, Mo, Sb, Zr modulate the electronicstructure [46]. The quinary electrocatalysts were synthesizedusing the reported protocols [38]. In our model, the molar ratioof elements in the quinary catalysts was fixed at 1:1:1:1:1. Atotal of 3003 candidate catalysts (5 elements from 15 elements)were generated using a combinatorial enumeration approach. Inorder to construct a chemical composition search space, only theelemental composition was considered as the variable (Figure 1a).To construct a training dataset, 10 quinary electrocatalystswere randomly selected from the 15 elements, followed byeNO3RR, eNO2RR, and HER linear sweep voltammetry (LSV),and eNO3RR controlled potential electrolyses (CPEs) at −0.3 Vversus RHE. NO3RR products were quantified using colorimetricmethods [47, 48]. All electrocatalysts exhibit higher currentresponses toward NO3RR and NO2RR than toward HER (FigureS1). During the eNO3RR, the measured total current density con-sists of two contributions: the reduction current originating fromNO3− itself and the current associated with the HER. Therefore,we performed mathematical integration of the eNO3RR and HERLSV curves and calculated their difference to quantify the neteNO3RR current (denoted as 𝑗𝑁𝑂3𝑅𝑅 − 𝑗𝐻𝐸𝑅, Figure 1b), whichserves as an indicator of the net eNO3RR current, followingthe same principle for the eNO2RR versus HER (𝑗𝑁𝑂2𝑅𝑅 − 𝑗𝐻𝐸𝑅,Figure 1c). The 𝑗𝑁𝑂3𝑅𝑅 − 𝑗𝐻𝐸𝑅 and 𝑗𝑁𝑂2𝑅𝑅 − 𝑗𝐻𝐸𝑅 of the trainingelectrocatalysts were calculated and plotted against NH3 yieldsAngewandte Chemie International Edition, 2026ive Commons LicenseFIGURE 1 (a) Schematic illustration of HMC applied to NO3RR electrocatalyst screening. By employing dimensionally reduced reactiondescriptors, the eNO3RR-to-NH3 pathway can be effectively extracted from the multi-pathway eNO3RR reaction network. Schematic illustrations of(b) 𝑗𝑁𝑂3𝑅𝑅 − 𝑗𝐻𝐸𝑅 and (c) 𝑗𝑁𝑂2𝑅𝑅 − 𝑗𝐻𝐸𝑅 calculations from eNO3RR, eNO2RR, and HER LSV profiles. (d) Correlations of 𝑗𝑁𝑂3𝑅𝑅 − 𝑗𝐻𝐸𝑅 with 𝑌𝑁𝐻3and(e) 𝑗𝑁𝑂2𝑅𝑅 − 𝑗𝐻𝐸𝑅 with 𝑌𝑁𝐻3in the initial training dataset. (f) Plots of electrocatalyst number against 𝑗𝑁𝑂2𝑅𝑅 − 𝑗𝐻𝐸𝑅 , and (g) plots of 𝑗𝑁𝑂2𝑅𝑅 − 𝑗𝐻𝐸𝑅against 𝑌𝑁𝐻3after 10 “training-suggestion-experiment” loops.(bcodTavA 15213773, 2026, 25, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/anie.2469199 by National Institute For, Wiley Online Library on [18/06/2026]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creativ𝑌𝑁𝐻3) from CPEs (Figure 1d,e). There is no clear correlationetween 𝑗𝑁𝑂3𝑅𝑅 − 𝑗𝐻𝐸𝑅 and 𝑌𝑁𝐻3(Figure 1d), but a strong linearorrelation is observed between 𝑗𝑁𝑂2𝑅𝑅 − 𝑗𝐻𝐸𝑅 and 𝑌𝑁𝐻3(R2 valuef 0.917) (Figure 1e), suggesting that 𝑗𝑁𝑂2𝑅𝑅 − 𝑗𝐻𝐸𝑅 is a reliableescriptor for the NO3RR electrocatalyst.he electrocatalyst compositions and 𝑗𝑁𝑂2𝑅𝑅 − 𝑗𝐻𝐸𝑅 were useds descriptors for HMC to predict promising electrocatalystsia a statistical model and optimization algorithms. In eachngewandte Chemie International Edition, 2026loop, HMC gave two suggestions followed by the synthesisand electrochemical measurements of eNO2RR, HER LSVs andeNO3RR CPEs at −0.3 V. The newly obtained compositions and𝑗𝑁𝑂2𝑅𝑅 − 𝑗𝐻𝐸𝑅 were incorporated into the dataset for the next loopand plotted against 𝑌𝑁𝐻3to assess the correlation (Figure S2).After the first loop, the predicted electrocatalysts show a stronglinear correlation between 𝑗𝑁𝑂2𝑅𝑅 − 𝑗𝐻𝐸𝑅 and 𝑌𝑁𝐻3(R2 value of0.872) (Figure S2a), further confirming the high accuracy of thedescriptors. At the fourth loop, the material consists of Fe, Co, Ni,3 of 11e Commons LicenseFIGURE 2 (a) Products distribution and NH3 production rates of FeCoNiCuGa during eNO3RR CPEs in 0.1 M KOH + 1 M KNO3. (b) The jtotal and𝑗𝑁𝐻3from eNO3RR CPEs on FeCoNiCuGa. (c) Products distribution and 𝑗𝑁𝐻3of NO3RR CPEs on FeCoNiCuGa, Pt, and Rh at −0.3 V. (d) NO3RR and (e)NO2RR LSV curves of FeCoNiCuGa, Pt, and Rh in 0.1 M KOH + 1 M KNO3 and 0.1 M KOH + 0.2 M KNO2. (f) Elemental occurrence frequencies acrossthe full HMC dataset. Elemental occurrence frequencies in (g) the top 10 electrocatalysts ranked by 𝑌𝑁𝐻3, and (h) the top 10 electrocatalysts ranked by𝑗𝑁𝑂2𝑅𝑅 − 𝑗𝐻𝐸𝑅 . (i) Products distribution and 𝑗𝑁𝐻3of NO3RR CPEs on Ag-containing electrocatalysts at −0.3 V.Cw𝑗cSttwsaat2TdFcor4 15213773, 2026, 25, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/anie.2469199 by National Institute For, Wiley Online Library on [18/06/2026]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creativu, Ga (FeCoNiCuGa) has emerged as the optimal electrocatalystith a highest 𝑌𝑁𝐻3of 1.43 mmol. In subsequent loops, although𝑁𝑂2𝑅𝑅 − 𝑗𝐻𝐸𝑅 exhibits an upward trend, the suggested electro-atalysts show lower NH3 yields than FeCoNiCuGa (Figures 1f,3). This performance saturation prompted the termination ofhe screening after 10 loops. In total, 30 electrocatalysts (10 fromhe initial training dataset and 20 suggested electrocatalysts)ere synthesized, accounting for 1% of the 3003 candidates. Atrong correlation between 𝑗𝑁𝑂2𝑅𝑅 − 𝑗𝐻𝐸𝑅 and 𝑌𝑁𝐻3maintainsfter 10 loops (Figure 1g), confirming that HMC effectivelyccelerates the electrocatalyst screening and significantly reduceshe experimental workload..2 Data-Driven and Mechanical Analysishe eNO3RR CPEs from −0.1 to −0.6 V versus RHE were con-ucted by using FeCoNiCuGa as the model system (Figure 2a,b).eCoNiCuGa exhibits high NH3 Faradaic efficiencies (FEs), totalurrent densities (jtotal) and NH3 partial current densities (𝑗𝑁𝐻3)ver a wide potential range and the maximum NH3 productionate of 18.1 mmol mgcat−1 is achieved at −0.6 V (Figure 2a).of 11At −0.3 V, the NH3 production rate reaches at 9.8 mmolmgcat−1, outperforming platinum group metals (PGMs) such asPt and Rh, showing its potential for large-scale implementation(Figure 2c). Rh has a high NO2RR response (Figure 2e), butinferior eNO3RR performance (Figure 2d) limits its NH3 FE.HER was revealed to be the dominant reaction on Pt, suggestingits inferior performance as a NO3RR electrocatalyst (FigureS4). A 12-hour CPE was conducted at −0.3 V to evaluate thedurability of the FeCoNiCuGa (Figure S5). The current densityremained almost constant throughout continuous operation andthe NH3 FE showed minimal variation over time, indicatingstable electrocatalytic activity and selectivity of FeCoNiCuGa.The absence of any significant performance decay suggests thatno functional degradation occurred. To determine electrocatalystcompositions, the structural analyses including x-ray diffractom-etry (XRD), scanning transmission electron microscopy (STEM),and x-ray photoelectron spectroscopy (XPS) of FeCoNiCuGa wereperformed (Figures S6−S8). An obvious electrocatalyst layer canbe observed in the cross-sectional high-angle annular dark field(HAADF) STEM image of FeCoNiCuGa (Figure S7a). The EDSmapping of O, Fe, Co, Ni, Cu, and Ga elements show the uniformdistribution, further confirming the even composition of FeCoN-Angewandte Chemie International Edition, 2026e Commons LicenseiaGitTwpofafra6ttwcl0aNpAtApFaF0risrtecTttFdNFwNpmeatmFslaiA 15213773, 2026, 25, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/anie.2469199 by National Institute For, Wiley Online Library on [18/06/2026]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable CreatiCuGa (Figures S7b−g). The XPS analysis reveals that Fe, Co, Ni,nd Cu exist as the mixed valence states, while Ga exists as thea(III) state in FeCoNiCuGa (Figure S8). FeCoNiCuGa displaysn an amorphous matrix, where the synergistic interaction amonghe elements has a coordination effect in NH3 formation.o identify the high electrochemical properties of FeCoNiCuGa,e studied the correlations between the electrocatalyst com-osition and both the 𝑗𝑁𝑂2𝑅𝑅 − 𝑗𝐻𝐸𝑅 and 𝑌𝑁𝐻3. The elementalccurrence frequencies of the total HMC dataset were countedirst. Within this dataset, the Fe, Co, Cu, Ni, and Ga elementsppear 22, 19, 19, 16, and 15 times, respectively, ranking the topive elements (Figure 2f). Analyses of the top 10 electrocatalystsanked by 𝑌𝑁𝐻3and 𝑗𝑁𝑂2𝑅𝑅 − 𝑗𝐻𝐸𝑅 show that Fe, Cu, Ni, Co,nd Ga consistently appear with frequencies of 9, 9, 8, 8, and, and 10, 8, 8, 6, and 6, respectively (Figure 2g,h), showinghe importance in NH3 production. Notably, Ag appears fourimes in the top 10 𝑗𝑁𝑂2𝑅𝑅 − 𝑗𝐻𝐸𝑅 but demonstrates no associationith the high 𝑌𝑁𝐻3(Figure 2g,h, Table S1). Excluding Ag-ontaining electrocatalysts from the HMC dataset improves theinear correlation between 𝑗𝑁𝑂2𝑅𝑅 − 𝑗𝐻𝐸𝑅 and 𝑌𝑁𝐻3(R2 value of.774) (Figure S9). The NO3RR LSVs of Ag-containing electrocat-lysts show onset potentials around 0.2 V (Figure S10) and highO2− FEs upon CPEs at −0.3 V (Figure 2i), suggesting that Agromotes eNO3RR to NO2−. These results indicate that, despiteg-containing electrocatalysts having high eNO2RR activities,he NO2− accumulation ultimately constrains the NH3 yields.iming to study the Ag component effect on the electrochemicalroperties, we decided to add Ag into FeCoNiCuGa to prepareeCoNiCuGaAg and check its eNO3RR and eNO2RR LSVs,nd CPE at −0.3 V (Figure 3a−c). Compared to FeCoNiCuGa,eCoNiCuGaAg exhibits a positive NO3RR onset potential near.25 V but lower current densities in the negative potentialegion (Figure 3a). However, the eNO2RR LSVs of FeCoN-CuGa and FeCoNiCuGaAg exhibit a similar trend (Figure 3b),uggesting that these two materials have comparable apparenteaction kinetics toward eNO2RR. To investigate the origin ofhe electrocatalytic performance of both electrocatalysts, thelectrochemically active surface area (ECSA) measurements wereonducted for FeCoNiCuGa and FeCoNiCuGaAg (Figure S11).he FeCoNiCuGaAg demonstrates higher double-layer capaci-ance (Cdl) than FeCoNiCuGa, indicating a higher ECSA. Despitehis higher ECSA, FeCoNiCuGaAg shows a lower NH3 FE thaneCoNiCuGa (Figure 3c), excluding surface area effects as theominant factor. Meanwhile, the introduction of Ag shifts theO3RR onset potential positively and leads to a higher NO2−E, suggesting weakened *NO2− adsorption on FeCoNiCuGaAg,hich consequently suppresses NH3 formation. Therefore, theH3 selectivity is governed by intrinsic active-site properties,articularly the variations in *NO2− binding strength, rather thanorphology or specific surface area effects. Subsequently, thelemental occurrence frequencies of the bottom three electrocat-lysts ranked by 𝑗𝑁𝑂2𝑅𝑅 − 𝑗𝐻𝐸𝑅 (also the bottom three 𝑌𝑁𝐻3elec-rocatalysts) were analyzed, showing that both Zr and Mo appearost frequently (Figure S12, Table S1). Thus, FeCoNiCuGaZr,eCoNiCuGaMo, and FeCoNiCuGaZrMo were synthesized andubsequently subjected to electrochemical measurements to ana-yze the degradation issue (Figure 3a−c). Mo addition causesslight decline in both NO3RR and NO2RR LSVs of FeCoN-CuGaMo, while FeCoNiCuGaZr and FeCoNiCuGaZrMo exhibitngewandte Chemie International Edition, 2026vinferior NO3RR and NO2RR performance (Figure 3a−c), demon-strating that Zr introduction leads to significant performancedegradation (Figure 3a,b). Zr adding impedes eNO2RR, therebynegatively affecting NH3 yields (Figure 3c). Summarizing thestudy on elemental effects, the results from FeCoNiCuGaAgand FeCoNiCuGaZr electrochemical measurements with theprevious observations on Ag-containing materials and the HMCdescriptors, *NO2− formation and its conversion to subsequentsteps are critical in determining NH3 production.The analysis based on HMC descriptors and elemental roles ofAg and Zr identified *NO2−/NO2− as the critical intermediatesto NH3 yields. Thus, elucidating *NO2− behavior is essen-tial for understanding the eNO3RR mechanism within MEPTframework. Given that multi-component electrocatalysts couldalter product distribution, the individual contributions of eachcomponent were investigated to better understand *NO2− effect.Upon CPE, single-metal materials demonstrate significantlylower NH3 FE and 𝑗𝑁𝐻3than FeCoNiCuGa (Figure 3e). NO2−is the dominant product on Cu and Ni whereas Co favorsNH3 formation, implying that Cu and Ni accelerate *NO2−formation with a limited *NO2− conversion effect while Copromotes *NO2− reduction. Additionally, FeNiCuGaAg exhibitsa diminished NO2RR response and a higher NO2− FE thanFeCoNiCuGaAg (Figures S13−S15), further confirming that theCo facilitates the eNO2RR process to elevate NH3 productivity.Meanwhile, Fe contributes to limited NH3 FE; Ga exhibits thelowest 𝑗𝑁𝐻3(Figure 3e). A comparative analysis of FeCoNiCu-GaAg and FeCoNiCuAg clarified that although FeCoNiCuAg hashigher NO3RR and NO2RR responses than FeCoNiCuGaAg, thedominant product is NO2− (Figures S13−S15). As NO3RR is aMEPT with 2-electron/proton transfer HER as a side reaction, theelectrons/protons’ behavior within HER could significantly affecteNO3RR performance. FeCoNiCuAg demonstrates higher HERperformance than FeCoNiCuGaAg (Figure S16), suggesting anintense proton transfer competition on FeCoNiCuAg during NH3formation. Ga has been identified as a promoter: introducing Gamakes NO3RR intermediates protonation more preferential thanHER. Thus, Ga assists in selective NH3 formation. To further vali-date the *NO2− effect, four-metal materials without Cu, Ni, or Ga(FeCoNiGa, FeCoCuGa, FeCoNiCu) were synthesized, followedby electrochemical measurements. All electrocatalysts exhibitlower eNO3RR performance than FeCoNiCuGa, but similarNO2RR trends (Figures 3d, S17). Upon CPEs, all electrocatalystshave comparable NH3 FE but lower 𝑗𝑁𝐻3than that of FeCoN-iCuGa (Figure 3e). The product current density is determinedby the total current density and FE, which reflects the productformation rate [49]. CPE and eNO3RR LSV results infer that theabsence of Cu, Ni, or Ga hinders the NO3RR to *NO2− (Figure 3e).Therefore, despite all electrocatalysts maintaining comparableeNO2RR activity to FeCoNiCuGa, inferior *NO2− generationlimits NH3 production rate. The eNO3RR Tafel analysis onFeCoNiCuGa toward NH3 formation reveals a slope of 47 mVdec−1, close to the theoretical value of 59 mV dec−1, indicating thatthere is one electron transfer before rate-determining step (RDS)[50] and suggesting that NH3 production is governed by *NO2−formation (Figure 3f). The HMC descriptor shows that NH3productivity depends not only on NO2− conversion but also HERsuppression, clarifying the importance of both eNO2RR and HERto overall performance. The HER Tafel slope on FeCoNiCuGais 145 mV dec−1 (Figure S18), indicating *H formation is the5 of 11e Commons LicenseFIGURE 3 (a) eNO3RR, and (b) eNO2RR LSVs of FeCoNiCuGa, FeCoNiCuGaAg, FeCoNiCuGaZr, FeCoNiCuGaMo, and FeCoNiCuGaZrMo in0.1 M KOH + 1 M KNO3 and 0.1 M KOH + 0.2 M KNO2. (c) Products distribution and 𝑗𝑁𝐻3of FeCoNiCuGa, FeCoNiCuGaAg, FeCoNiCuGaZr,FeCoNiCuGaMo, and FeCoNiCuGaZrMo from NO3RR CPE at −0.3 V in 0.1 M KOH + 1 M KNO3. (d) NO3RR LSVs of FeCoNiCuGa, FeCoNiCu,FeCoCuGa, and FeCoNiGa. (e) Products distribution and 𝑗𝑁𝐻3of Fe, Ga, Ni, Cu, Co, FeCoNiCuGa, FeCoNiCu, FeCoCuGa, and FeCoNiGa from NO3RRCPE at −0.3 V in 0.1 M KOH + 1 M KNO3. (f) eNO3RR and (g) eNO2RR Tafel slope for NH3 formation on FeCoNiCuGa. (h) The schematic illustrationof NO3RR with different *NO2− adsorption.R9Naiotd*fTeaipwbS6 15213773, 2026, 25, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/anie.2469199 by National Institute For, Wiley Online Library on [18/06/2026]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable CreatDS [51, 52]. The FeCoNiCuGa shows an eNO2RR Tafel slope of7 mV dec−1 (Figure 3g), implying that eNO2RR is restricted byO2− adsorption. *NO2− competes with *H reducing *H cover-ge, inhibiting HER while promoting NH3 formation. However,nadequate *NO2− coverage causes low NH3 productivity. Basedn the Sabatier principle [53–55] and above results, we concludehat *NO2− adsorption strength and its subsequent reductionominate NH3 productivity toward eNO3RR (Figure 3h). StrongNO2− binding enables high *NO2− coverage facilitating NH3ormation while weak *NO2− binding favors NO2− formation.o further support the role of *NO2− in NH3 formation, RRDExperiments were conducted on FeCoNiCuGa, FeCoNiCuGaAg,nd FeCoNiCuGaZr. It is well known that the RRDE techniques a powerful electrochemical approach for studying kinetics,rovided there is an effective mathematical model. In this study,e develop a new eNO3RR kinetic model for the RRDE techniqueased on a previous ORR RRDE model (Figure 4a, details inupporting Information). In our model, k1 represents the directof 118-electron reduction of NO3− to NH3 without *NO2− process; k2represents a 2-electron NO3− reduction to *NO2− process whilek−2 represents the reversed path; k3 is the deeper reductionof *NO2− to NH3; k5 and k6 represent *NO2− desorption andNO2− adsorption processes, respectively. Upon analyzing the𝑁𝐼𝐷𝐼𝑅𝜔−0.5 plots at different potentials, the tandem electrode anddirect reduction processes can be distinguished. The Pt0.5PdAu-coated Pt ring was used as the ring electrode and catalyzedan efficient NO2− oxidation reaction (NO2OR), even at 10 mMNO2− concentration (Figure S19). Using the FeCoNiCuGa-coatedPt disk as the disk electrode, the collection efficiency (N) wascalculated in 0.1 M KOH + 10 mM KNO3 (Figure S20, TableS2). The average value (0.408) was adopted for subsequentcalculations (Table S2). The RRDE setup was used to investigatethe eNO3RR and NO2OR behaviors of FeCoNiCuGa at variousrotation rates in 0.1 M KOH + 1 M KNO3 (Figure 4b). DuringRRDE analysis, the ring potential was 1.675 V versus RHE, atwhich the oxygen evolution reaction (OER) current was relativelysmall (Figure S19). Within the potential range of 0.25–0.30 VAngewandte Chemie International Edition, 2026ive Commons LicenseFIGURE 4 (a) Schematic illustration of eNO3RR RRDE model (* represents adsorption state; s represents surface; a represents bulk solution).(b) eNO3RR LSVs and corresponding ring current of FeCoNiCuGa at different rotation rates in 0.1 M KOH + 1 M KNO3. (c) Plots of 𝑁𝐼𝐷𝐼𝑅−ω−0.5 forFeCoNiCuGa at −0.075 to −0.225 V versus RHE potential range. (d) The J − S plots derived from the J and S values of 𝑁𝐼𝐷𝐼𝑅−ω−0.5 plots at differentpotentials. In situ FT-IR spectra of NO3RR on (e) FeCoNiCuGa, (f) FeCoNiCuGaAg, and (g) FeCoNiGa. (h) In situ FT-IR spectra of eNO2RR onFeCoNiCuGa. (i) Schematic illustration of the proposed eNO3RR reaction pathway.voacbatIranopoU−tA 15213773, 2026, 25, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/anie.2469199 by National Institute For, Wiley Online Library on [18/06/2026]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicableersus RHE (Figure 4b), the NO3RR had not yet been initiatedn the disk electrode. The corresponding ring current exhibitedstable plateau, which mainly originated from the OER andould therefore be regarded as the background current. Thisackground contribution was subtracted in the subsequent datanalysis to obtain a more accurate net ring current associated withhe oxidation of reaction intermediates (details in Supportingnformation). Both NO3RR and NO2OR currents increase withotation rates, indicating elevated mass transfer. The 𝑁𝐼𝐷𝐼𝑅−ω−0.5nalysis was conducted from −0.075 to −0.225 V, where HER isegligible (Figure S21). From −0.075 to −0.225 V (with an intervalf 0.025 V), the intercepts (J) obtained from 𝑁𝐼𝐷𝐼𝑅−ω−0.5 plots at allotentials are greater than 1 (Figure 4c), suggesting the presencef the k3 process (𝐽 = 1 + 4𝑘3𝑘5, details in Supporting Information).sing J and the slopes (S) at different potentials to draw JS plots, the intercept (J′) is 1.02 (Figure 4d), demonstratinghat the k process (direct NO RR to NH ) is negligible on1 3 3ngewandte Chemie International Edition, 2026 CFeCoNiCuGa (𝐽′ = 1 + 4𝑘1𝑘2, details in Supporting Information).Thus, the eNO3RR on FeCoNiCuGa proceeds via consecutiveprocesses of NO3RR to *NO2− and *NO2− to NH3. The RRDEanalysis was also applied to FeCoNiCuGaZr, which exhibits lowerJ values than FeCoNiCuGa but follows the same consecutivepathway (Figures S22−S23). At a specific potential, the J valueof the 𝑁𝐼𝐷𝐼𝑅−ω−0.5 curve can be used to quantify the tendencyof the k3 to the k5 process ( 𝑘3𝑘5), signifying the importance of*NO2− to final NH3 productivity. The J values of FeCoNiCuGa,FeCoNiCuGaZr, and FeCoNiCuGaAg at −0.2 V were 1.48, 1.12,and 1.17, respectively (Figures 4c, S23, and S24). Compared toFeCoNiCuGaAg and FeCoNiCuGaZr, FeCoNiCuGa has a low*NO2− desorption (k5) rate and prefers k3 process, which isadvantageous for NH3 formation. In contrast, in FeCoNiCuGaAgand FeCoNiCuGaZr, a suppressed k3 process or an accelerated k5process results in limited NH3 production. The RRDE analysissuccessfully excludes the existence of the direct eNO RR to NH3 37 of 11reative Commons LicensepNpTtC(cpi(rAaast(AFg**(N1oiTGNotpaaio(lNoossCk*ssstib[tiF8 15213773, 2026, 25, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/anie.2469199 by National Institute For, Wiley Online Library on [18/06/2026]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creativrocess and identifies the importance of *NO2− conversion toH3 productivity in the tandem electrode process, providing aowerful reference for subsequent mechanism studies.he overall pathway was determined using in situ Fourierransform infrared (FT-IR) spectra of eNO3RR and eNO2RR.ompared to the in situ FT-IR spectra of HER on FeCoNiCuGaFigure S25), the downward peaks at 1365 cm−1 during NO3RRorrespond to NO3− consumption (Figure 4e) [56]. The depletioneaks at 1631 and ∼3300 cm−1 indicate H2O consumption, whichs accompanied by considerable NO3RR intermediates formationFigures 4e, S26) [43, 57, 58]. The upward peaks at 1441 cm−1epresent the N─H vibration, indicating NH3 formation [59, 60].t 0.05 V, the *NO2− bands at 1230 cm−1 appear [56, 61, 62] andttenuate from −0.1 V to −0.2 V. Correspondingly, new bandst 1118 cm−1 relating to N─O in *NH2OH emerge [61, 63–65],uggesting *NO2− conversion. A downward peak correspondingo NO3− consumption emerges at 0.2 V on FeCoNiCuGaAgFigure 4f), which is consistent with the NO3RR LSV results.dditionally, H2O depletion bands at 1630 cm−1 are observed oneCoNiCuGaAg (Figure 4f) [43]. The *NO2− peaks at 1239 cm−1radually strengthen from 0.1 to −0.3 V, suggesting sluggishNO2− conversion [56, 61, 62]. To further confirm the role ofNO2−, NO3RR in situ FT-IR spectra were recorded on FeCoNiGaFigure 4g). The absence of Cu hinders NO3− conversion toO2−. Accordingly, the NO3− and H2O consumption bands at356 [56] and 1635 cm−[143] and N−H vibration bands at 1442 cm−1nly appear at negative potentials [59, 60], demonstrating thatnadequate *NO2− formation restricts NH3 yields.he eNO2RR in situ FT-IR spectra of FeCoNiCuGa, FeCoNiCu-aAg, and FeCoNiCuGaZr were used to explore the eNO2RR toH3 pathway. In Figure 4h, the downward peaks at 1230 cm−1n FeCoNiCuGa correspond to NO2− consumption, whereashe increased peaks around 1290 cm−1 are the characteristiceaks of *NH2 [66]. The upward trends at 1540, 1434, 1160,nd 1110 cm−1 indicate *NO [67], NH3 [59, 60], *NH2 [60],nd *NH2OH formation [61, 63–65], confirming the evolution ofntermediates. Furthermore, the eNO2RR in situ FT-IR spectraf FeCoNiCuGaAg are comparable to those of FeCoNiCuGaFigure S27). In contrast, Zr impedes the eNO2RR process, thusimiting *NO2− transformation on FeCoNiCuGaZr. As a result,O2− consumption can only be observed at negative potentialsn FeCoNiCuGaZr (Figure S28). Additionally, the high Tafel slopef FeCoNiCuGaZr toward the NO2RR (152 mV dec−1, Figure S29)uggests that Zr weakens *NO2− adsorption, thereby reducingurface coverage and negatively impacting NH3 yield.orrelating in situ FT-IR spectra with data analysis and theinetic analyses based on RRDE unveils the significance ofNO2− in NH3 generation. A high *NO2− formation rate andtrong *NO2− adsorption ensure sufficient *NO2− coverage andubsequent transformation. In alkaline conditions, the electrodeurface is typically negatively charged, which can result in elec-rostatic repulsion toward anions such as NO3−/NO2−. However,on distributions within the electric double layer (EDL) cannote described solely by classical electrostatic considerations alone68]. During electrolysis, the high concentration K+ cations couldune the local electric field near the electrode surface, whichs benefit to the negatively charged species stabilization [69].urthermore, the continuous NO3−/NO2− consumption on theof 11surface leads to a local concentration gradient, generating a diffu-sion flux toward the electrocatalyst surface. The combination ofmigration and diffusion enables the efficient transport of anionsto the reaction plane in EDL. Moreover, the structural disorder ofthe amorphous matrix may result in a more heterogeneous localelectric field distribution at the interface, potentially influencingthe interfacial water structure. This interfacial modulation maycontribute to the suppression of competing HER thus facilitatingNO3RR to NH3. Based on these findings, we conclude thatthe overall NH3 formation proceeds via the 𝑁𝑂−3→∗ 𝑁𝑂−2→∗𝑁𝑂 →∗ 𝑁𝐻2𝑂𝐻 →∗ 𝑁𝐻2 → 𝑁𝐻3 pathway (Figure 4i).3 ConclusionIn this study, based on the concept of direct simplification ofMEPT, 𝑗𝑁𝑂2𝑅𝑅 − 𝑗𝐻𝐸𝑅 is strategically employed as a descriptor inBO-based HMC to accelerate eNO3RR electrocatalysts screening.A highly efficient FeCoNiCuGa electrocatalyst with an NH3production rate of 9.8 mmol mgcat−1 at −0.3 V is identified injust 1% of the trials across the entire candidate space. Our dataanalysis and mechanistic investigations including RRDE and insitu FT-IR spectra reveal that NH3 productivity is governed by*NO2− intermediates. Weak *NO2− binding leads to low *NO2−coverage, resulting in low NH3 productivity. Conversely, strong*NO2− binding enables high surface *NO2− coverage, ensuringNH3 productivity. Currently, AI platforms for MEPT reactions arerestricted by the complex reaction process and limited data avail-ability. This work provides a paradigm for identifying descriptorswithin complex reaction networks by incorporating key reactionsteps into a human−machine collaborative workflow. Underalkaline conditions, HER kinetics are significantly suppressed,NO2− behaviors emerge as the dominant step that governs theoverall reaction rate, which provides an experimentally accessiblebasis for constructing the descriptor. Despite the pronouncedchanges in the interfacial local environments, electron donors,and electric double layer structures under neutral or acidicconditions, the same dimensional reduced principle can stillguide the design of new descriptors. This work successfullydecouples the NO3RR process and integrating AI platformsinto the MEPT framework and clearly elucidates the crucialintermediates evolution at the solid−liquid interface, establishinga mechanistic foundation for future electrocatalytic applications.Author ContributionsYingying Cheng: methodology, data curation, investigation, validation,formal analysis, visualization, writing – original draft, writing – reviewand editing. Masaki Takeguchi: methodology, data curation, inves-tigation, validation, formal analysis, visualization. Abraham CastroGarcia: methodology, data curation, investigation, validation, formalanalysis, visualization, writing – review and editing. Ken Sakaushi:conceptualization, methodology, data curation, investigation, validation,formal analysis, supervision, funding acquisition, visualization, projectadministration, resources, writing – review and editing, writing – originaldraft.AcknowledgementsThis work was supported by the MEXT Program: Data Creation andUtilization-Type Material Research and Development Project Grant Num-Angewandte Chemie International Edition, 2026e Commons LicensebaEiCTDTfpRT2H3tQp14fP5RC06SePo7o(8RE09ESt01oMh1bZCa1SOt 1.2131628.A 15213773, 2026, 25, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/anie.2469199 by National Institute For, Wiley Online Library on [18/06/2026]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creatier JPMXP1122712807, and Japan Science and Technology Agency (JST)s part of Adopting Sustainable Partnerships for Innovative Researchcosystem (ASPIRE), Grant Number JPMJAP2421. The authors arendebted to the National Institute for Material Science.onflicts of Interesthe authors declare no conflicts of interest.ata Availability Statementhe data that support the findings of this study are available on requestrom the corresponding author. The data are not publicly available due torivacy or ethical restrictions.eferences1. R. Gurney, “The Quantum Mechanics of Electrolysis,” Proceedings ofhe Royal Society of London Series A 134 (1931): 137–154.. R. Dogonadze, A. Kuznetsov, and V. Levich, “Quantum Theory ofydrogen Overvoltage,” Elektrokhim 3 (1967): 739–742.. W. Schmickler, “The Influence of the Inner Solvation Sphere on Elec-rochemical Outer Sphere Redox Reactions of Transition Metal Ions—Auantum Mechanical Approach,” Berichte der Bunsengesellschaft fürhysikalische Chemie 77 (1973): 991–994, https://doi.org/10.1002/bbpc.9730771048.. A. Soudackov and S. Hammes-Schiffer, “Multistate Continuum Theoryor Multiple Charge Transfer Reactions in Solution,” Journal of Chemicalhysics 111 (1999): 4672–4687, https://doi.org/10.1063/1.479229.. M. T. M. Koper, “Thermodynamic Theory of Multi-electron Transfereactions: Implications for Electrocatalysis,” Journal of Electroanalyticalhemistry 660 (2011): 254–260, https://doi.org/10.1016/j.jelechem.2010.10.04.. K. Sakaushi, T. Kumeda, S. Hammes-Schiffer, M. M. Melander, and O.ugino, “Advances and Challenges for Experiment and Theory for Multi-lectron Multi-proton Transfer at Electrified Solid–liquid Interfaces,”hysical Chemistry Chemical Physics 22 (2020): 19401–19442, https://doi.rg/10.1039/D0CP02741C.. R. R. Adžić, S. Strbac, and N. Anastasijević, “Electrocatalysis of Oxygenn Single Crystal Gold Electrodes,” Materials Chemistry and Physics 221989): 349–375.. H. Kita, S. Ye, and Y. Gao, “Mass Transfer Effect in Hydrogen Evolutioneaction on Pt Single-Crystal Electrodes in Acid Solution,” Journal oflectroanalytical Chemistry 334 (1992): 351–357, https://doi.org/10.1016/022-0728(92)80583-P.. N. M. Marković, R. R. Adžić, B. Cahan, and E. Yeager, “Structuralffects in Electrocatalysis: Oxygen Reduction on Platinum Low Indexingle-Crystal Surfaces in Perchloric Acid Solutions,” Journal of Elec-roanalytical Chemistry 377 (1994): 249–259, https://doi.org/10.1016/0022-728(94)03467-2.0. R. Jinnouchi and A. B. Anderson, “Electronic Structure Calculationsf Liquid-solid Interfaces: Combination of Density Functional Theory andodified Poisson-Boltzmann Theory,” Physical Review B 77 (2008): 245417,ttps://doi.org/10.1103/PhysRevB.77.245417.1. K. Ojha, N. Arulmozhi, D. Aranzales, and M. T. M. Koper, “Dou-le Layer at the Pt(111)–Aqueous Electrolyte Interface: Potential ofero Charge and Anomalous Gouy–Chapman Screening,” Angewandtehemie International Edition 59 (2020): 711–715, https://doi.org/10.1002/nie.201911929.2. T. Kumeda, L. Laverdure, K. Honkala, M. M. Melander, and K.akaushi, “Cations Determine the Mechanism and Selectivity of Alkalinexygen Reduction Reaction on Pt(111),” Angewandte Chemie Interna-ional Edition 62 (2023): e202312841.ngewandte Chemie International Edition, 202613. Y. Yang, R. G. Agarwal, P. Hutchison, et al., “Inverse Kinetic IsotopeEffects in the Oxygen Reduction Reaction at Platinum Single Crystals,”Nature Chemistry 15 (2023): 271–277.14. F. Jiao and B. Xu, “Electrochemical Ammonia Synthesis and Ammo-nia Fuel Cells,” Advanced Materials 31 (2019): e1805173, https://doi.org/10.1002/adma.201805173.15. B. H. Ko, B. Hasa, H. Shin, Y. Zhao, and F. Jiao, “ElectrochemicalReduction of Gaseous Nitrogen Oxides on Transition Metals at Ambi-ent Conditions,” Journal of the American Chemical Society 144 (2022):1258–1266, https://doi.org/10.1021/jacs.1c10535.16. International Renewable Energy Agency, Ammonia Energy Asso-ciation. IRENA Innovation Outlook Renewable Ammonia. can befound under: https://www.irena.org/publications/2022/May/Innovation-Outlook-Renewable-Ammonia, 2022 (accessed: 19 September 2024).17. S. Han, H. Li, T. Li, et al., “Ultralow Overpotential Nitrate Reduction toAmmonia via a Three-step Relay Mechanism,” Nature Catalysis 6 (2023):402–414, https://doi.org/10.1038/s41929-023-00951-2.18. K. Dong, Y. Yao, H. Li, et al., “H2O2-Mediated Electrosynthesis ofNitrate From Air,” Nature Synthesis 3 (2024): 763–773, https://doi.org/10.1038/s44160-024-00522-8.19. C. Chen, X. Zhu, X. Wen, et al., “Coupling N2 and CO2 in H2O toSynthesize Urea Under Ambient Conditions,” Nature Chemistry 12 (2020):717–724, https://doi.org/10.1038/s41557-020-0481-9.20. P. Li, R. Li, Y. Liu, M. Xie, Z. Jin, and G. Yu, “Pulsed Nitrate-to-Ammonia Electroreduction Facilitated by Tandem Catalysis of NitriteIntermediates,” Journal of the American Chemical Society 145 (2023):6471–6479, https://doi.org/10.1021/jacs.3c00334.21. S. Zhu, K. Liu, Z. Feng, H. Jiang, and J. Lin, “The Dual Active SiteNi3Sn2-NiSnOx Alloy-Oxide Catalysts via Sn-Modulated Ni Coordinationfor Efficient Ammonia Synthesis,” Nano Research Energy 4 (2025):e9120188, https://doi.org/10.26599/NRE.2025.9120188.22. H. Zhang, H. Wang, X. Cao, et al., “Unveiling Cutting-Edge Devel-opments in Electrocatalytic Nitrate-to-Ammonia Conversion,” AdvancedMaterials 36 (2024): e2312746, https://doi.org/10.1002/adma.202312746.23. J. Liang, Z. Li, L. Zhang, et al., “Advances in Ammonia Electrosyn-thesis From Ambient Nitrate/Nitrite Reduction,” Chemistry 9 (2023):1768–1827, https://doi.org/10.1016/j.chempr.2023.05.037.24. X. Fan, C. Liu, X. He, et al., “Efficient Electrochemical Co-Reductionof Carbon Dioxide and Nitrate to Urea With High Faradaic Efficiencyon Cobalt-Based Dual-Sites,” Advanced Materials 36 (2024): e2401221,https://doi.org/10.1002/adma.202401221.25. Z. Y. Wu, M. Karamad, X. Yong, et al., “Electrochemical AmmoniaSynthesis via Nitrate Reduction on Fe Single Atom Catalyst,” Nature Com-munications 12 (2021): 2870, https://doi.org/10.1038/s41467-021-23115-x.26. K. B. Prater and A. J. Bard, “Rotating Ring-Disk Electrodes III.Catalytic and ECE Reactions,” Journal of the Electrochemical Society 117(1970): 1517–1520, https://doi.org/10.1149/1.2407362.27. A. Damjanovic, M. A. Genshaw, and J. O. M. Bockris, “DistinctionBetween Intermediates Produced in Main and Side Electrodic Reactions,”Journal of Chemical Physics 45 (1966): 4057–4059, https://doi.org/10.1063/1.1727457.28. H. S. Wroblowa, P. Yen Chi, and G. Razumney, “Electroreductionof Oxygen a New Mechanistic Criterion,” Journal of ElectroanalyticalChemistry and Interfacial Electrochemistry 69 (1976): 195–201, https://doi.org/10.1016/S0022-0728(76)80250-1.29. A. J. Appleby and M. Savy, “Kinetics of Oxygen Reduction ReactionsInvolving Catalytic Decomposition of Hydrogen Peroxide,” Journal ofElectroanalytical Chemistry and Interfacial Electrochemistry 92 (1978):15–30, https://doi.org/10.1016/S0022-0728(78)80113-2.30. R. W. Zurilla, R. K. Sen, and E. Yeager, “The Kinetics of theOxygen Reduction Reaction on Gold in Alkaline Solution,” Journal ofthe Electrochemical Society 125 (1978): 1103–1109, https://doi.org/10.1149/9 of 11ve Commons Licensehttps://doi.org/10.1002/bbpc.19730771048https://doi.org/10.1063/1.479229https://doi.org/10.1016/j.jelechem.2010.10.004https://doi.org/10.1039/D0CP02741Chttps://doi.org/10.1016/0022-0728(92)80583-Phttps://doi.org/10.1016/0022-0728(94)03467-2https://doi.org/10.1103/PhysRevB.77.245417https://doi.org/10.1002/anie.201911929https://doi.org/10.1002/adma.201805173https://doi.org/10.1021/jacs.1c10535https://www.irena.org/publications/2022/May/Innovation-Outlook-Renewable-Ammoniahttps://doi.org/10.1038/s41929-023-00951-2https://doi.org/10.1038/s44160-024-00522-8https://doi.org/10.1038/s41557-020-0481-9https://doi.org/10.1021/jacs.3c00334https://doi.org/10.26599/NRE.2025.9120188https://doi.org/10.1002/adma.202312746https://doi.org/10.1016/j.chempr.2023.05.037https://doi.org/10.1002/adma.202401221https://doi.org/10.1038/s41467-021-23115-xhttps://doi.org/10.1149/1.2407362https://doi.org/10.1063/1.1727457https://doi.org/10.1016/S0022-0728(76)80250-1https://doi.org/10.1016/S0022-0728(78)80113-2https://doi.org/10.1149/1.21316283i(3oS(3oR(3RFC03oi13HC43EA3lE(3oA14RCD4Nh4NM14MA4c(4Rte4BN04Am1 15213773, 2026, 25, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/anie.2469199 by National Institute For, Wiley Online Library on [18/06/2026]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Cre1. J. C. Huang, R. K. Sen, and E. Yeager, “Oxygen Reduction on Platinumn 85% Orthophosphoric Acid,” Journal of the Electrochemical Society 1261979): 786–792, https://doi.org/10.1149/1.2129139.2. X. Xing, D. A. Scherson, and C. Mak, “The Electrocatalytic Reductionf Nitrate Mediated by Underpotential-Deposited Cadmium on Gold andilver Electrodes in Acid Media,” Journal of the Electrochemical Society 1371990): 2166–2175, https://doi.org/10.1149/1.2086905.3. N. M. Marković, H. A. Gasteiger, and P. N. Ross, “Oxygen Reductionn Platinum Low-Index Single-Crystal Surfaces in Alkaline Solution:otating Ring DiskPt(hkl) Studies,” Journal of Physical Chemistry 1001996): 6715–6721, https://doi.org/10.1021/jp9533382.4. U. A. Paulus, T. J. Schmidt, H. A. Gasteiger, and R. J. Behm, “Oxygeneduction on a High-Surface Area Pt/Vulcan Carbon Catalyst: A Thin-ilm Rotating Ring-Disk Electrode Study,” Journal of Electroanalyticalhemistry 495 (2001): 134–145, https://doi.org/10.1016/S0022-0728(00)0407-1.5. Y. Chen, H. Zhu, M. Rasmussen, and D. Scherson, “Rational Designf Electrocatalytic Interfaces: The Multielectron Reduction of Nitraten Aqueous Electrolytes,” Journal of Physical Chemistry Letters 1 (2010):907–1911, https://doi.org/10.1021/jz1005253.6. Z. W. Chen, J. Li, P. Ou, et al., “Unusual Sabatier Principle onigh Entropy Alloy Catalysts for Hydrogen Evolution Reactions,” Natureommunications 15 (2024): 359, https://doi.org/10.1038/s41467-023-44261-.7. D. Yin, B. Li, B. Gao, et al., “Overcoming Energy-Scaling Barriers:fficient Ammonia Electrosynthesis on High-Entropy Alloy Catalysts,”dvanced Materials 37 (2025): e2415739.8. K. Sakaushi, W. Hoisang, and R. Tamura, “Human–Machine Col-aboration for Accelerated Discovery of Promising Oxygen Evolutionlectrocatalysts With on-Demand Elements,” ACS Central Science 92023): 2216–2224, https://doi.org/10.1021/acscentsci.3c01009.9. S. Nie, Y. Xiang, L. Wu, et al., “Active Learning Guided Discoveryf High Entropy Oxides Featuring High H2-production,” Journal of themerican Chemical Society 146 (2024): 29325–29334, https://doi.org/10.021/jacs.4c06272.0. H. Xu, Y. Ma, J. Chen, W. X. Zhang, and J. Yang, “Electrocatalyticeduction of Nitrate—A Step towards a Sustainable Nitrogen Cycle,”hemical Society Reviews 51 (2022): 2710–2758, https://doi.org/10.1039/1CS00857A.1. K. Fan, W. Xie, J. Li, et al., “Active Hydrogen Boosts Electrochemicalitrate Reduction to Ammonia,” Nature Communications 13 (2022): 7958,ttps://doi.org/10.1038/s41467-022-35664-w.2. E. Murphy, Y. Liu, I. Matanovic, et al., “Elucidating Electrochemicalitrate and Nitrite Reduction Over Atomically-dispersed Transitionetal Sites,” Nature Communications 14 (2023): 4554, https://doi.org/10.038/s41467-023-40174-4.3. S. Liang, X. Teng, H. Xu, L. Chen, and J. Shi, “H* Species Regulation byn-Co(OH)2 for Efficient Nitrate Electro-Reduction in Neutral Solution,”ngewandte Chemie International Edition 63 (2024): e202400206.4. A. Bagger, W. Ju, A. S. Varela, P. Strasser, and J. Rossmeisl, “Electro-hemical CO2 Reduction: A Classification Problem,” Chemphyschem 182017): 3266–3273, https://doi.org/10.1002/cphc.201700736.5. X. Ouyang, W. Qiao, Y. Yang, et al., “Intensifying Interfacialeverse Hydrogen Spillover for Boosted Electrocatalytic Nitrate Reduc-ion to Ammonia,” Angewandte Chemie International Edition 64 (2025):202422585.6. Q. Yang, W. Xu, S. Gong, et al., “Atomically Dispersed Lewis Acid Sitesoost 2-electron Oxygen Reduction Activity of Carbon-based Catalysts,”ature Communications 11 (2020): 5478, https://doi.org/10.1038/s41467-20-19309-4.7. K.-H. Kim, H. Lee, X. Huang, et al., “Energy-Efficient Electrochemicalmmonia Production From Dilute Nitrate Solution,” Energy & Environ-ental Science 16 (2023): 663–672, https://doi.org/10.1039/D2EE03461A.0 of 1148. J. Yu, R. T. Gao, X. Guo, N. Truong Nguyen, L. Wu, and L. Wang,“Electrochemical Nitrate Reduction to Ammonia on AuCu Single-AtomAlloy Aerogels Under Wide Potential Window,” Angewandte ChemieInternational Edition 64 (2024): e202415975.49. Y. Liu and C. C. L. McCrory, “Modulating the Mechanism of Elec-trocatalytic CO2 Reduction by Cobalt Phthalocyanine Through PolymerCoordination and Encapsulation,” Nature Communications 10 (2019):1683, https://doi.org/10.1038/s41467-019-09626-8.50. K. Sakaushi, “Quantum Electrocatalysts: Theoretical Picture, Elec-trochemical Kinetic Isotope Effect Analysis, and Conjecture to Under-stand Microscopic Mechanisms,” Physical Chemistry Chemical Physics 22(2020): 11219–11243, https://doi.org/10.1039/D0CP01052A.51. M. T. M. Koper, “Analysis of Electrocatalytic Reaction Schemes: Dis-tinction Between Rate-determining and Potential-Determining Steps,”Journal of Solid State Electrochemistry 17 (2012): 339–344, https://doi.org/10.1007/s10008-012-1918-x.52. H. Prats and K. Chan, “The Determination of the HOR/HER ReactionMechanism From Experimental Kinetic Data,” Physical Chemistry Chem-ical Physics 23 (2021): 27150–27158, https://doi.org/10.1039/D1CP04134G.53. J. K. Nørskov, T. Bligaard, A. Logadottir, et al., “Trends in the ExchangeCurrent for Hydrogen Evolution,” Journal of the Electrochemical Society152 (2005): J23–J26.54. J. Greeley, T. F. Jaramillo, J. Bonde, I. Chorkendorff, and J. K. Nørskov,“Computational High-throughput Screening of Electrocatalytic Materialsfor Hydrogen Evolution,” Nature Materials 5 (2006): 909–913, https://doi.org/10.1038/nmat1752.55. Z. W. Seh, J. Kibsgaard, C. F. Dickens, I. Chorkendorff, J. K. Nørskov,and T. F. Jaramillo, “Combining Theory and Experiment in Electrocataly-sis: Insights Into Materials Design,” Science 355 (2017): eaad4998, https://doi.org/10.1126/science.aad4998.56. E. Pérez-Gallent, M. C. Figueiredo, I. Katsounaros, and M. T. M.Koper, “Electrocatalytic Reduction of Nitrate on Copper Single Crystalsin Acidic and Alkaline Solutions,” Electrochimica Acta 227 (2017): 77–84,https://doi.org/10.1016/j.electacta.2016.12.147.57. S. Liu, T. Qian, M. Wang, et al., “Proton-Filtering Covalent OrganicFrameworks With Superior Nitrogen Penetration Flux Promote AmbientAmmonia Synthesis,” Nature Catalysis 4 (2021): 322–331, https://doi.org/10.1038/s41929-021-00599-w.58. Y. Wan, M. Pei, Y. Tang, et al., “Interfacial Water Regulation for NitrateElectroreduction to Ammonia at Ultralow Overpotentials,” AdvancedMaterials 37 (2025): e2417696, https://doi.org/10.1002/adma.202417696.59. Y. Zhang, H. Zheng, K. Zhou, et al., “Conjugated CoordinationPolymer as a New Platform for Efficient and Selective Electroreductionof Nitrate Into Ammonia,” Advanced Materials 35 (2023): e2209855.60. S. Lu, G. Lin, H. Yan, et al., “In Situ Facet Transformation Engineeringover Co3O4 for Highly Efficient Electroreduction of Nitrate to Ammonia,”ACS Catalysis 14 (2024): 14887–14894.61. J. Y. Fang, Q. Z. Zheng, Y. Y. Lou, et al., “Ampere-Level Current DensityAmmonia Electrochemical Synthesis Using CuCo Nanosheets Simulat-ing Nitrite Reductase Bifunctional Nature,” Nature Communications 13(2022): 7899, https://doi.org/10.1038/s41467-022-35533-6.62. X. Huang, Y. Li, S. Xie, et al., “The Tandem Nitrate and CO2Reduction for Urea Electrosynthesis: Role of Surface N-Intermediates inCO2 Capture and Activation,” Angewandte Chemie International Edition63 (2024): e202403980, https://doi.org/10.1002/anie.202403980.63. Y. Zhou, W. Zhang, P. Guo, et al., “Continuous Regulation of Cu Elec-tronic States by Rectifying Schottky Contacts Enhancing ElectrochemicalNitrate Reduction to Ammonia,” Inorganic Chemistry Frontiers 11 (2024):3503–3510, https://doi.org/10.1039/D4QI00614C.64. Y. Tang, Z. Jiang, Y. Yuan, et al., “Selective Electrosynthesis ofHydroxylamine From Aqueous Nitrate/Nitrite by Suppressing FurtherReduction,” Nature Communications 15 (2024): 9800, https://doi.org/10.1038/s41467-024-54204-2.Angewandte Chemie International Edition, 2026ative Commons Licensehttps://doi.org/10.1149/1.2129139https://doi.org/10.1149/1.2086905https://doi.org/10.1021/jp9533382https://doi.org/10.1016/S0022-0728(00)00407-1https://doi.org/10.1021/jz1005253https://doi.org/10.1038/s41467-023-44261-4https://doi.org/10.1021/acscentsci.3c01009https://doi.org/10.1021/jacs.4c06272https://doi.org/10.1039/D1CS00857Ahttps://doi.org/10.1038/s41467-022-35664-whttps://doi.org/10.1038/s41467-023-40174-4https://doi.org/10.1002/cphc.201700736https://doi.org/10.1038/s41467-020-19309-4https://doi.org/10.1039/D2EE03461Ahttps://doi.org/10.1038/s41467-019-09626-8https://doi.org/10.1039/D0CP01052Ahttps://doi.org/10.1007/s10008-012-1918-xhttps://doi.org/10.1039/D1CP04134Ghttps://doi.org/10.1038/nmat1752https://doi.org/10.1126/science.aad4998https://doi.org/10.1016/j.electacta.2016.12.147https://doi.org/10.1038/s41929-021-00599-whttps://doi.org/10.1002/adma.202417696https://doi.org/10.1038/s41467-022-35533-6https://doi.org/10.1002/anie.202403980https://doi.org/10.1039/D4QI00614Chttps://doi.org/10.1038/s41467-024-54204-26EI6tB(6Nm26iC16tEI2SAISA 15213773, 2026, 25, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/anie.2469199 by National Institute For, Wiley Online Libra5. Y. Liu, Z. Zhuang, Y. Liu, et al., “Shear-Strained Pd Single-Atomlectrocatalysts for Nitrate Reduction to Ammonia,” Angewandte Chemienternational Edition 63 (2024): e202411396.6. K. Yang, S. H. Han, C. Cheng, C. Guo, T. Li, and Y. Yu, “Unveilinghe Reaction Mechanism of Nitrate Reduction to Ammonia over Cobalt-ased Electrocatalysts,” Journal of the American Chemical Society 1462024): 12976–12983, https://doi.org/10.1021/jacs.3c13517.7. W. Liao, J. Wang, G. Ni, et al., “Sustainable Conversion of Alkalineitrate to Ammonia at Activities Greater than 2 A cm−2,” Nature Com-unications 15 (2024): 1264, https://doi.org/10.1038/s41467-024-45534-.8. P. Li, Y. Jiao, J. Huang, and S. Chen, “Electric Double Layer Effectsn Electrocatalysis: Insights From Ab Initio Simulation and Hierarchicalontinuum Modeling,” JACS Au 3 (2023): 2640–2659, https://doi.org/10.021/jacsau.3c00410.9. W. Wen, S. Fang, Y. Zhou, Y. Zhao, P. Li, and X. Y. Yu, “Modulatinghe Electrolyte Microenvironment in Electrical Double Layer for Boostinglectrocatalytic Nitrate Reduction to Ammonia,” Angewandte Chemienternational Edition 63 (2024): e202408382, https://doi.org/10.1002/anie.02408382.upporting Informationdditional supporting information can be found online in the Supportingnformation section.upporting File: anie72478-sup-0001-SuppMat.docx.ngewandte Chemie International Edition, 2026 11 of 11ry on [18/06/2026]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons Licensehttps://doi.org/10.1021/jacs.3c13517https://doi.org/10.1038/s41467-024-45534-2https://doi.org/10.1021/jacsau.3c00410https://doi.org/10.1002/anie.202408382 Accelerated Discovery-to-Unveiling of High-Performance and Affordable Ammonia Electrode Process by Human-Machine Collaboration Framework 1 | Introduction 2 | Results and Discussion 2.1 | Descriptor Establishment 2.2 | Data-Driven and Mechanical Analysis 3 | Conclusion Author Contributions Acknowledgements Conflicts of Interest Data Availability Statement References Supporting Information